Longitudinal changes in leukocyte telomere length and mortality in elderly Swedish men

Introduction

Telomeres are structures at the end regions of chromosomes in eukaryotic organisms. They consist of repeated, species specific, sequences of hexanucleotides, which partly serve to prevent unlimited cell replication and to protect the cell from genome instability [1]. At birth telomere length (TL) is hereditary and varies significantly between individuals [2]. During life, there is a balance between telomere shortening and telomere extension, but TL is typically shortened with each cell cycle (due to the end-replication problem) yielding shorter telomeres with aging [3]. Environmental and lifestyle factors have been identified to influence telomere shortening [4-6], but their role in TL dynamics remains elusive.

TL has been proposed as a promising biological clock of aging [7], and strong relationships to age-related diseases have been described [8,9]. Many studies have also reported shorter telomeres to be associated with increased mortality [10-21], whilst others have not [22-27]. There are few longitudinal studies on TL and cause-specific mortality in the literature. To our knowledge, there are only three studies based on healthy individuals that have assessed longitudinal changes in TL in relation to mortality. Two of these studies found no relationship between telomere attrition and mortality [22,25], and the third study found a relationship only to cardiovascular mortality [13]. The general absence of relationship in those studies may be explained by the age of study participants and short follow-up. The mean ages at baseline in the first two studies were 90 and 45 years, respectively, and changes in TL might be a poor predictor of mortality at those ages. The third study, which found a relationship to cardiovascular mortality but not to total mortality, had a limited follow-up period of only 2.5 years. Study population and sample size, together with e.g. measurement techniques, represents likely caveats that may hamper possibilities to capture TL attrition as a predictor of mortality.

We hypothesized that short telomeres and especially shortening of telomeres could be related to all-cause and cause-specific mortality in a population-based study of elderly individuals. Therefore, data from the ULSAM (Uppsala Longitudinal Study of Adult Men) cohort was used to assess whether TL and longitudinal changes in TL were related to all-cause, cardiovascular, and cancer mortality in elderly Swedish men.

Results

Baseline characteristics of our 247 participants examined at age 71 and 81 are depicted in Table 1, revealing a rather healthy set of individuals being only slightly obese, with few smokers and being relatively physically active. During follow-up, the study group became less active and smoking habits improved, but they remained relatively weight stable. At age 81, a majority of the participants had pharmacological treatment and most of them had received cardiovascular (49%) or cancer (21%) diagnosis. On average, TL decreased by 30% in the study group, 83% displayed TL attrition and 10% extended their telomeres, defined as more than 15% loss or extension in TL between examinations. Furthermore, a strong relationship between baseline TL and loss in TL was found (β=0.76, p<0.0001 for absolute values); i.e. longer baseline telomeres was related to a higher loss in TL. During the mean follow-up time of 7.4 years, using the 81-year-examination as starting-point, 178 (72%) deaths occurred, of which 65 cases (37%) were due to cardiovascular disease and 61 (34%) were due to cancer.

Discussion

In the present study we investigated whether baseline TL at both age 71 and 81, as well as longitudinal changes in TL, were related to mortality in elderly men. In our population, baseline TL was strongly associated with mortality, with a 40 to 70% increased risk of all-cause mortality, and over 2-fold increased risk of cancer mortality among individuals with short telomeres. The corresponding risk relations for cardiovascular mortality were more inconsistent, indicating that short telomeres at age 71, but not 81, may enhance the risk of future cardiovascular mortality. For longitudinal changes, TL attrition was apparent in an absolute majority of our individuals (83%), but 10% actually extended their telomeres. Extension of telomeres have previously been described in the literature [28], although some claim that this is an effect of measurement error [29]. However, we find measurement error an unlikely explanation to TL extension in our study group considering the long time-span between examinations and a limited CV in our qPCR assays. Our risk analyses for longitudinal changes revealed that TL attrition did not predict all-cause or cancer mortality, but we found a 60% decreased risk for cardiovascular mortality in those who shortened their telomeres. This unexpected finding remained robust after controlling for influential confounders such as baseline TL, pronounced cardiovascular risk factors, and previous cardiovascular diagnosis.

Overall, our data indicate increased risk of mortality with short telomeres at baseline, a relationship that was especially strong when using TL-data derived from the 71-year-examination. This is in line with most previous studies [10-21], but not all [22-27], describing a negative relationship between baseline TL and mortality. Previous studies also indicate that this relation becomes weaker with increasing age [30]. The latter is coherent with our observations, in which risk relations were strong when baseline was set to 71 years, but not when using the 81-year-examination. A weakened relationship with increasing age might be anticipated considering the loss in TL that occurs in most individuals, and especially in those who start with long telomeres. The distribution of TL in our individuals at age 71 was wider than at age 81, indicating that our population became more homogenic in terms of TL as they grew older. This would in turn limit the possibilities to detect risk associations. Still, we found a pronounced relation between short telomeres and future cancer mortality, independent of choice of baseline. This relation became especially strong after adjusting for potential confounders, such as prevalent cancer disease. Previous studies on TL and cancer mortality reveal conflicting results [9], and whether TL has a causative role remains to be established. It is likely that the relation may differ for different cancer forms and further being highly dependent on follow-up time, age of study participants, and possibilities to control for influential confounders. Today, most attention within this field is given to the role of telomerase, an enzyme that is responsible for elongating telomeres. The ability to maintain an adequate TL, by upregulating telomerase activity, has been proposed as a prerequisite for tumor development, but current data suggest a more complex scenario [9].

We further analyzed whether longitudinal changes in TL could predict all-cause or cause-specific mortality. This was also tested in a sub-group of individuals who had short telomeres at baseline and displayed TL attrition between the two examinations. The rationale for the latter approach relates to a study by Epel et al, in which extremely high 12-year mortality was found among men with short TL at baseline, when comparing those who experienced TL shortening with those who did not [13]. Interestingly, we were not able to replicate this, with overall null findings. TL attrition in this sub-group was neither related to all-cause, cardiovascular nor cancer mortality. The discrepancy between these two studies is hard to explain, but supposedly statistical power could play a role. Although the size of our sub-group was limited (n=82) it was higher than in the aforementioned study, which should limit the risk of chance findings. However, corresponding analyses in the whole study group, comparing those who experienced TL attrition (n=206) with those who did not (n=41), revealed some interesting results. After controlling for essential confounders, TL attrition did neither predict all-cause nor cancer mortality. However, individuals who experienced TL attrition had a lower risk of future cardiovascular mortality than those who maintained or lengthened their telomeres. This risk relation remained after controlling for influential confounders and indicated a 60% risk reduction of dying from cardiovascular disease. The notion that shortening of telomeres would protect against cardiovascular mortality is novel and intriguing, and available data on longitudinal changes in TL are not coherent with our findings. On the contrary, a recent study on patients with stable coronary heart disease found decreased mortality in patient who lengthened their telomeres [31]. Moreover, a newly published study on lifelong changes in TL found a more pronounced TL shortening among CVD patients compared with non-CVD patients [32]. Finally, a study on breast cancer survivors found increased overall mortality risk (cardiovascular mortality was not assessed) in those who shortened their telomeres [33]. It is noteworthy, though, that the two aforementioned mortality studies were based on individuals with an established diagnosis, and similar studies in healthy subjects do not convincingly show any increased risk of mortality with TL attrition. One study found an elevated risk of cardiovascular death among men who shortened their telomeres [13], but those results may have been affected by a short follow-up (2.5y) and missing covariates. Another study in the oldest old (~90y at baseline) could not detect any relation between TL attrition and future mortality [25], and one can argue that shortening of telomeres may be less important in this age group. A third study, with a large study sample, found no relation to total mortality for changes in TL over 10 years [22]. The time-span for assessing TL changes was similar to ours, but the participants were much younger (mean 45y at baseline). Still, those results are coherent with our data on all-cause mortality and it would have been interesting to know the corresponding results for cardiovascular mortality in that study.

Despite strengths of this study, such as well-characterized participants, valid follow-up time for the age group and minimal loss to follow-up, there were several limitations. First, we had a relatively limited sample size. This forced us to combine deaths from coronary heart disease and cerebrovascular disease, which may have affected our risk relations to cardiovascular mortality. It also prevented us from analyzing mortality from specific cancer forms, and since prostate cancer represents a major part of cancer deaths in our material (45%) we cannot exclude the possibility that those associations are driven by prostate cancer per se. Second, the possible influence of selection bias cannot be ignored since our participants at age 71 were healthy and survival up to the 81-year-examination was a prerequisite to be included in these analyzes. Third, technical issues such as precision of methodology and DNA quality may have influenced our results. However, we have no indication of poor DNA quality or exceptionally poor precision in our PCR method that would make our results unreliable. The length of our telomeres, though, may oscillate over time, as also suggested in the literature [28,34]. This would limit the possibilities of measuring longitudinal changes in TL with great precision. Finally, we cannot exclude the possibilities of residual confounding, and neither the possibility that some of our selected covariates may not be true confounders, which could yield distorted risk estimates. Still, risk estimates in our model with such covariates (i.e. the fully adjusted model), did not deviate substantially from the corresponding risk estimates in the minimally adjusted model.

In conclusion, our study confirms previous publications showing an overall shortening of TL length over time, and that TL extension occurs in a subset of individuals. Our data also indicate an increased risk of all-cause and cancer mortality, but not clearly so for cardiovascular mortality, in elderly men having short telomeres at baseline. Longitudinal changes were only related to cardiovascular mortality, with a pronounced risk reduction in those who shortened their telomeres. The latter result is intriguing and should be interpreted with some caution until additional studies may confirm such a relation.

Materials and Methods

Supplementary Materials

Supplementary Figure